Re: [Condor-users] Master won't start

[Erik Paulson <epaulson@xxxxxxxxxxx>]

On Wed, Mar 08, 2006 at 04:23:37PM -0500, Steve Huston wrote:
> We ran Condor awhile ago and turned it off while we did some changes,
> and now I'm setting it up again.  I've got a Fedora 3 machine (quad
> Opteron) which I'm setting up with Condor 6.7.17, and every time I run
> it I get the following error:
> 
>  Neither the environment variable CONDOR_CONFIG,
>  /etc/condor/, nor ~condor/ contain a condor_config file.
>  Either set CONDOR_CONFIG to point to a valid config file,
>  or put a "condor_config" file in /etc/condor or ~condor/
> 
> There *is* a condor_config file in ~condor/ and this is the same setup
> we used before with the previous machines (though I edited the file so
> it reflects the new setup, and new options since this is also a newer
> version of the software).  If I do as it says and set an environment
> variable, I instead get the following:
> 
>  ERROR: the uid specified in CONDOR_IDS config file variable (9016)
>  does not exist in your password information.
>  Please set CONDOR_IDS to the '.' seperated uid, gid pair that
>  should be used by condor.
> 
> Our authentication is done via LDAP, but the UID 9016 does in fact match
> the condor user.  All the files are owned by the condor user as well.
> 

Is your config file in NFS? Do you have root squash turned on?

> If I instead run the daemon as root (which is what I want to do), I get
> the "no config" error no matter what, even if I set the environment
> variable - *except* if I try to run it through gdb, then I get the "no
> UID" error.  I hoped that there'd be some debugging in the binaries,
> which is how I found the difference running through gdb.
> 
> I've attached an strace of the process if it will help someone, but I
> don't see where it even tries to look up anything to determine that
> there is no file in ~condor, nor does it look in /etc/condor.  This
> trace was when running it as root.
> 

The master forks and goes into the background, and the process that
looks for the config file is the forked child. Run the master with
the '-f' flag:

strace condor_master -f

to get a better trace.

-Erik